Bankroll Management for Betting: Units, Staking Plans and Kelly

Bankroll management is the part of betting that decides how much you stake, not what you bet on. It sounds dull next to picking winners, but it does two jobs nothing else does. It stops a bad run from wiping you out, and it makes your records comparable, so you can tell whether your betting actually works.

None of it creates an edge. A good staking plan applied to bets with negative expected value only loses more slowly. What it does is stop a sound method from being ruined by bet sizing, and stop a losing run from turning into chasing. This guide covers how to set a bankroll, what a betting unit is, the three main staking methods with worked numbers, and what drawdowns realistically look like.

Setting a betting bankroll

A bankroll is a sum of money set aside only for betting. The rules are simple, and most of the damage happens when people bend them:

• It is money you can afford to lose entirely without touching bills, rent, savings or debt repayments.
• It is kept apart from everyday money. Many bettors use a separate bank account or e-wallet so the balance is always visible.
• It is one pot, even if it is spread across several bookmaker accounts. Add the balances together and treat the total as the bankroll.
• You decide in advance whether and when you add to it. Topping it up after a loss so you can keep betting is chasing, not bankroll management.

The size is a personal decision. £500 is a sensible bankroll, and so is £100. What matters is that the number is fixed before you start and that losing it would be disappointing rather than damaging. If you find you cannot stick to the amount, our responsible gambling page covers deposit limits, time-outs and where to get support.

Betting units: what they are and how big they should be

A betting unit is your standard stake, expressed as a share of the bankroll. Most bettors set one unit at 1% to 2% of the starting bankroll, so a £500 bankroll gives a unit of £5 to £10 and a $2,000 bankroll gives $20 to $40.

Units do two things. They keep stakes small enough to survive normal losing runs, which are bigger than most people think. And they let you describe results without the amounts: +12 units over 400 bets means the same thing for someone staking £5 and someone staking $100. Return on investment (ROI) is profit divided by total staked. If you placed 300 one-unit bets and finished +9.6 units, your ROI is 9.6 ÷ 300 = 3.2%.

Be wary of unit claims from tipsters. A record quoted in units means little if the unit size changes from bet to bet, if big "five-unit plays" are remembered when they win and forgotten when they lose, or if the quoted odds were never available to most people. Your own records should be stricter than that.

Flat staking and percentage staking

Most staking plans start from one of two simple rules: stake the same amount every time, or stake the same share of whatever the bankroll is now. Both work with small units. They differ in how they behave after a run of losses or wins.

Flat staking: the same stake every time

Flat staking means every bet is the same size, usually one unit, whatever the odds or your confidence. With a £1,000 bankroll and a 2% unit, every bet is £20 until you choose to review the unit.

It is the simplest method and the hardest to misuse. A losing run costs a predictable amount: ten losers in a row at £20 is £200, or 20% of the starting bankroll. It also keeps records clean, because every bet counts equally.

The drawback is that stakes do not respond to the bankroll. If you are down 40%, a £20 stake is now 3.3% of what is left, so your risk per bet has risen without you deciding it should. If you are up 50%, your stakes are relatively smaller than you planned. Many bettors handle this by keeping flat stakes but recalculating the unit at fixed points, such as the first of each month or every 100 bets, and never in the middle of a losing run.

Percentage staking: a share of the current bankroll

Percentage staking sets each stake as a fixed share of the current bankroll, not the starting one. At 2%, a £1,000 bankroll stakes £20. After a bad week takes it to £800, the stake is £16. After a good month takes it to £1,200, the stake is £24.

The benefit shows up in losing runs. Ten straight losers at 2% of the current bankroll leave £1,000 × 0.98^10 = £817.07, against £800 with flat £20 stakes, and in theory the bankroll never reaches zero. The cost is a slower recovery, because you climb back with smaller stakes than you fell with.

Percentage staking also has an asymmetry that surprises people. A 10% gain followed by a 10% loss does not take you back to where you started: 1.10 × 0.90 = 0.99, a 1% loss overall, and the same happens in the other order. Big swings cost you even when the wins and losses are equal in size, which is one more reason to keep each stake small.

The Kelly criterion and fractional Kelly

The Kelly criterion sizes each bet according to your edge. With decimal odds the formula is: stake as a share of bankroll = ((p × odds) - 1) ÷ (odds - 1), where p is your estimated probability of winning. The top of the fraction is your expected profit per unit staked; the bottom is the net odds. Our value betting guide covers how to estimate p in the first place.

Two examples. If you rate a bet at 2.00 as a 55% chance, Kelly says ((0.55 × 2.00) - 1) ÷ 1 = 10% of the bankroll. If you rate a 5.00 shot at 25%, Kelly says ((0.25 × 5.00) - 1) ÷ 4 = 6.25%. When your edge is zero or negative, Kelly says do not bet at all, which is a useful discipline in itself.

The catch is that Kelly assumes p is correct, and in sports betting it never is exactly. Here is what happens when it is wrong. You believe a bet at evens (2.00) has a 55% chance, so you stake the full Kelly 10% every time, but the true chance is 52%. The true Kelly stake was 4%. At 10% stakes your bankroll's long-run growth rate is about -0.10% per bet, even though every single bet has a positive expected value. Over 1,000 such bets the typical bankroll ends at roughly a third of where it started. At 8%, double the true Kelly, the growth rate is almost exactly zero. At 4% it is positive. Overbetting a real edge can still shrink your bankroll.

That is why most people who use Kelly stake a fraction of it. Half Kelly on the 55% example is 5% of the bankroll and quarter Kelly is 2.5%. A fraction gives up some growth if your estimates are right, in exchange for much smaller swings and far more protection if they are too optimistic. Kelly also assumes one bet at a time. If several bets run at once, cap your total exposure, for example at 10% of the bankroll, rather than staking a full fraction on each.

Drawdowns: what a normal losing run looks like

A drawdown is the fall from your highest bankroll to a later low point. Every bettor has them, including those with a real edge, and they are usually deeper than people expect.

Take a bettor who wins 53% of spread bets at -110, an edge of about 1.2% per bet, staking one unit on each of 500 bets. The expected profit is only about +5.9 units, while the standard deviation of the result is about 21 units, so luck is far bigger than the edge over that sample. We ran that scenario 100,000 times in a simulation:

• The typical (median) largest drawdown was about 21.5 units.
• One run in ten had a drawdown of 37 units or more.
• About 38% of runs finished the 500 bets with a loss, despite the edge.

The chance of a streak of eight or more straight losers somewhere in those 500 bets is about 47%. Bettors who prefer longer odds see longer streaks: at a 40% win rate, a run of ten or more straight losers turns up in about 70% of 500-bet sequences.

With a 1% unit, a 21.5-unit drawdown is a fall of more than a fifth of the bankroll. With a 5% unit and flat stakes, the same run of bets would wipe the bankroll out, which is the clearest argument for small units. Decide in advance what you will do in a drawdown. A sensible rule is to review your method, not your stakes, after a fixed fall such as 25%, and never to increase stakes to win the money back.

Record keeping: the numbers to track

Bankroll management only works if you know where the bankroll stands, so log every bet when you place it: date, sport, market, odds taken, stake in units and money, bookmaker, result and, where possible, the closing odds. A spreadsheet is enough.

From those columns you can track units won, ROI, average odds and your largest drawdown so far. Closing odds let you calculate closing line value, which shows far sooner than profit whether your prices are beating the market; the line movement and CLV guide explains how. Records also expose the cost of habits you might not notice. Someone who places a £5 four-leg acca every Saturday stakes £260 a year on bets where, as the accumulator guide shows, the margin compounds with every leg.

Common bankroll mistakes

The pattern behind most of them is letting the last result decide the next stake.

Chasing losses is the obvious one. Doubling up to recover yesterday's losers turns an ordinary drawdown into a disaster, and it is one of the recognised warning signs of a gambling problem. Raising stakes after a winning streak is the quieter version: a run of wins says little about your edge, but bigger stakes certainly mean bigger swings.

Confidence staking, where a "strong" pick gets five units and a weak one gets one, sounds sensible but often means the largest stakes go on the bets you were most excited about, which are not necessarily the ones with the biggest edge. If you want stakes to vary, tie them to a calculated edge, as Kelly does, not to a feeling. Finally, many bettors count withdrawn winnings as banked while leaving losses in the account, which makes the bankroll look healthier than it is. Count the whole pot every time.

Setting up your staking plan: a checklist

Before your next bet:

• Fix the bankroll at an amount you can afford to lose and write it down.
• Set one unit at 1% to 2% of it. If you mostly bet at long odds, stay nearer 1%.
• Choose flat or percentage staking. Use Kelly only if you put a probability on every bet, and then only a quarter or half of it.
• Cap total exposure on bets that are running at the same time.
• Set a review point, such as a 25% drawdown or every 200 bets, where you check the method rather than the stakes.
• Log every bet, including closing odds.
• Bet only with operators licensed where you live, since laws differ by country and, in the US and India, by state, and some Indian states ban online betting. Set deposit limits in your account as well.

A worked start: a £600 bankroll with a 1.5% unit gives £9 stakes. A bad week of six straight losers costs £54 and leaves £546, and the plan does not change. That is what the plan is for.